Low Illumination Image Decomposition and Details Enhancement Under Gradient Sparse and Least Square Constraint

HUANG Wei-guo; ZHANG Yong-ping; BI Wei; GAO Guan-qi; ZHU Zhong-kui

doi:10.3969/j.issn.0372-2112.2018.02.023

您当前的位置：

首页 >

文章列表页 >

Low Illumination Image Decomposition and Details Enhancement Under Gradient Sparse and Least Square Constraint

更新时间：2025-07-16

- Low Illumination Image Decomposition and Details Enhancement Under Gradient Sparse and Least Square Constraint
- Acta Electronica Sinica Vol. 46, Issue 2, Pages: 424-432(2018)
- 作者机构：
  
  苏州大学城市轨道交通学院,江苏,苏州,215131
- 作者简介：
- 基金信息：
  
  National Natural Science Foundation of China (No.51405320);Suzhou Natural Science Foundation of Jiangsu Province (No.SYG201511)
- DOI：10.3969/j.issn.0372-2112.2018.02.023
  CLC： TP391.4
- Published Online：25 February 2018，
  
  Published：2018
- 稿件说明：
移动端阅览
HUANG Wei-guo, ZHANG Yong-ping, BI Wei, et al. Low Illumination Image Decomposition and Details Enhancement Under Gradient Sparse and Least Square Constraint[J]. Acta Electronica Sinica, 2018, 46(2): 424-432.
DOI：

HUANG Wei-guo, ZHANG Yong-ping, BI Wei, et al. Low Illumination Image Decomposition and Details Enhancement Under Gradient Sparse and Least Square Constraint[J]. Acta Electronica Sinica, 2018, 46(2): 424-432. DOI： 10.3969/j.issn.0372-2112.2018.02.023.

摘要

低照度图像存在细节模糊、对比度低等问题.针对这些问题，本文提出一种低照度彩色图像增强算法.首先建立梯度稀疏和最小平方约束模型，将图像分解为结构层和细节层；然后采用提出的多尺度边缘保护细节增强算法强化图像的细节信息并滤波；最后把细节增强的图像经改进的Retinex算法映射，最终得到细节增强、亮度适宜、对比度较强的修复图像.实验结果表明，主观上：图像细节增强，亮度适宜；客观上：结构层图像的一维像素线性图显示其平滑特性效果较好，细节增强图的NIQE（5.5202）、BRISQE（31.1893）和PSNR（25.3625）特征较好，修复图像的熵值（7.4421）、边缘强度（128.3231）和平均亮度（121.1827）较好.本文算法实现了对低照度图像的有效分解及细节增强，并提高了图像综合质量.

Abstract

Low illumination images had the problems of fuzzy

low contrast and so on. In order to solve these problems

we put forward a low illumination image enhancement algorithm. Firstly

we established the gradient sparse and least square constraint model and decomposed the image into structure layer and detail layer. Then

the detail layer was enhanced by multi-scale edge-preserved algorithm and we used the Guided Filter to eliminate noise. Finally

the enhanced image was mapped by modified Retinex

we got the details enhanced

suitable brightness image. Experimental results show that performance is good

the 1D example figure of the contour is better than others

the figures of the details enhanced image NIQE(5.5202)

BRISQE(31.1893) and PSNR(25.3625) are better

the Entroy(7.4421)

Edge-Intensity(128.3231) and L-mean(121.1827) of the completed image are better as well. So the proposed algorithm shows a good performance in image enhancement.

关键词

Keywords

references

Views

283

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

No data

Related Author

No data

Related Institution

No data

⁰